Description
TitleExploiting network and dynamics for talent management
Date Created2021
Other Date2021-10 (degree)
Extent1 online resource (x, 111 pages) : illustrations
DescriptionNetwork and dynamics based analysis addresses various problems in talent management, emphasizing the connection, the influence and temporal factors in the working social network between different units (employees, departments, and companies). Nowadays, most contemporary commercial activities such as production, sales are organized by companies, which connect employees with various roles and levels into an intact network and involves broadly and deeply corporations; the environment composed of colleagues, departments, and companies inevitably generate positive or negative influences on the individuals, shaping their working behaviors, essentially impact the daily operations of the company; with the digitalization in industries, the abundant data of peoples' working make it feasible to analyze the role of social connections & influences in working and provide an insight into the patterns in communications between individual units in the working social network, and facilitate to develop data-driven applications to assist decision-making process in the management.
Our main contribution is to incorporate the social influence & network connections to provide intelligent solutions for different tasks in management such as turnover prediction and job position hierarchy extraction.Specifically, we first propose a contagious effect heterogeneous neural network (CEHNN) for individual turnover prediction. The network encodes the connections between employees based on the hierarchy of companies and formulates it as a sequential classification problem with the series of prior-connected-employee turnover events and other environment and profile factors. Further, we design a Turnover Influence-based Neural Network (TINN), which explicitly integrates the company's internal social networks like email and instant messages, models the diffusion of the turnover influences on these networks, and fuses them to forecast the turnover rate of departments. In addition, at the job level, we show that based on the aggregated job-hopping records, a framework is built to learn the vector representations of job positions across many companies, which is further utilized to reconstruct the hierarchy of job positions, identifying the potential links between them and predicting their rankings.The above approaches are validated for extensive experiments on real-world datasets, demonstrating their effectiveness, and the case studies reveal some interpretable findings which provide actionable insight into intelligent management.
NotePh.D.
NoteIncludes bibliographical references
Genretheses
LanguageEnglish
CollectionGraduate School - Newark Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.